Add LatinHypercube sampler as well as tests and a demonstrating notebook

[Samuel-Vangu]

Adds a LatinHypercube sampling class for integration on (hyper)cubic grids,
following up on the discussion in #322 .

What's included

• src/grid/latin_hypercube.py: LatinHypercube(n_points, dimension, seed=None, randomize=True, origin=None, axes=None). Stratifies every 1D marginal
  exactly (McKay, Beckman & Conover, 1979); randomize=False centers points
  in their stratum instead of jittering them. Supports mapping onto an
  arbitrary parallelepiped via origin/axes.
• src/grid/tests/test_latin_hypercube.py: 23 tests (stratification property,
  weights, domain mapping, integration accuracy, reproducibility)

• an example notebooks in examples/, per your earlier note about wanting a
  concrete use case: both compare LatinHypercube against baselines already
  in grid (plain Monte Carlo, Tensor1DGrids/Trapezoidal) on two
  chemistry-motivated integrands -- a sum of independent harmonic oscillators
  (additive) and a product of Gaussian-type functions (non-additive) -- to
  show honestly where LHS helps and where it doesn't.